Add benchmarks for BYTE_STREAM_SPLIT encoded Parquet FIXED_LEN_BYTE_ARRAY data

[etseidl]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
#6159 added support for using BYTE_STREAM_SPLIT with FIXED_LEN_BYTE_ARRAY primitive types. While some effort was put into optimizing the encoding path, the decoding path is largely unoptimized (and seemingly quite slow). It would be nice to have some benchmarks for the new encodings to guide future optimization efforts.

Describe the solution you'd like
Benchmarks for Float16/FIXED_LEN_BYTE_ARRAY(2) and DECIMAL/FIXED_LEN_BYTE_ARRAY(16) would be a good start for some likely to be used data types.

Describe alternatives you've considered

Additional context
See #6159 (comment) and following.

[alamb]

label_issue.py automatically added labels {'parquet'} from #6204